04:39:44 <elucidator> speaking of riscV https://www.reuters.com/technology/us-china-tech-war-risc-v-chip-technology-emerges-new-battleground-2023-10-06/ 
07:02:08 <Guest18> hi
07:02:51 <Guest18> will hashing goes up if we remove microcode from cpus
07:24:46 <sech1> If you remove microcode from cpus, you'll get nice paperweights instead of cpus
07:27:12 <moneromoooo> But fast paperweights ?
07:28:18 <moneromoooo> I think what was meant is, if Intel were to remove the programmability layer there to hardcode what it currently does.
07:28:43 <sech1> microcode is only used for complex instructions which are rarely used
07:31:30 <sech1> In RandomX, only CFROUND translates to instructions that use microcode
07:56:31 <Guest18> maybe microcode remove hidden specs from cpus
07:57:09 <Guest18> or better keeps hiden some specifics specs
07:57:19 <Guest18> sorry for my english
08:09:53 <elucidator> poor guy, if he thinks the microcode is the only way cpu manufacturer can put "hidden features" he'll be shocked when he learns manufacturer makes the hardware itself...
08:15:55 <Guest18> where motherboard manufacturer insert more code except cpus
08:16:29 <Guest18> we are not talking about servers that have management cpu
09:04:44 <moneromoooo> Various chips have "firmware" that get run on boot.
09:05:27 <moneromoooo> There's always the silicon that can have code that gets activated by seeing some trigger data (eg, network cards scanning for packets with some predefined bit pattern).
09:06:25 <moneromoooo> There are various "negative ring" layers, I forget now. The big one used to be the... EMM I think it was called ? Massive bunch of stuff your hypervisor doesn't see.
09:07:03 <moneromoooo> The "management engine" type stuff, which IIRC is different, but there's so much nowadays.
09:08:03 <moneromoooo> The baseband processor for phones. I totally dropped off that stuff ages ago, too much to keep track of. Basically, we're fucked in plemty of interesting ways.
09:08:48 <sech1> risc-v to the rescue :D
10:14:43 <Guest21> is libreboot the bets solution right now?
10:22:48 <elucidator> moneromoooo: like the Intel amt?
10:23:58 <Guest21> no i mean to run a computer without hidden code
10:24:53 <Guest21> https://github.com/corna/me_cleaner
10:25:06 <Guest21> this one is a good example
18:53:15 <m-relay> <4​rkal:matrix.org> Could a highly optimized Linux kernel improve hashrate? Is it worth my time to optimize?
18:58:25 <moneromoooo> Improving kernel performance is good regardless.
18:58:45 <moneromoooo> But to your question, I doubt it will make any difference at all.
19:15:02 <sech1> Maybe playing with spectre/meltdown and other mitigations can affect hashrate. Not just disabling, you can test both options
21:27:27 <hyc> a simple optimization could be just to delete support for all but the most fundamental of features. 
21:28:08 <hyc> a single type of storage device driver, a single specific network driver. boot up and run miner, nothing else.
21:29:03 <hyc> no dynamically loaded modules; dynamic linkage is a 5% overhead all on its own
21:32:38 <hyc> another approach - integrate miner into a unikernel. 
21:32:49 <hyc> no user mode context switches at all